HaochenW / ILSVRC2Yolo

Convert ILSVRC2015_VID .xml format to Yolo Darknet format

ILSVRC2Yolo

A format translator used for object detection label translation, which could translate ILSVRC2015_VID file into Yolo Darknet file.

Installation and Utilization

Clone

$ git clone https://github.com/HaochenW/ILSVRC2Yolo
$ cd ILSVRC2Yolo/

Direct utilization

$ python transxml.py --dataset_path your_ILSVRC_dataset_path

Required Parameters

1. --dataset_path

• ILSVRC_VID dataset path, for example:

  --dataset_path ./ILSVRC2015

2. --convert_output_path

• The output path after the conversion, for example, you could use:

• --convert_output_path ./ILSVRC2015/Annotation_yolo

3. --abs_path

• The output path of the image absolute path file, which is needed in training yolo model.

  --abs_path ./ILSVRC2015/Annotation_yolo

4. --cls_list_file

• The file that gives the name and number of the classes, the file is given in the folder

  --abs_path ./classes_ILSVRC.names

5. --img_type

• The type of the dataset image, in ILSVRC, it is .JPEG.

  --image_type .JPEG

Results

• The results will in the folder: "--convert_output_path"
• The folder nesting format will be the same as the folder in "Annotation", but the ".xml" format will be translated into ".txt' format

000465.txt

 ```bash
  22 0.438 0.497 0.496 0.994
 ```

valid.txt

```bash
/home/users/../code/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00138000/000247.JPEG
/home/users/../code/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00138000/000078.JPEG
/home/users/../code/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00138000/000075.JPEG
/home/users/../code/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00138000/000043.JPEG
/home/users/../code/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00138000/000013.JPEG
```